TennisEdge's AI model finds no betting value in A. Gea vs D. Shapovalov (Los Cabos on 2026-08-02) — the market is efficient, with D. Shapovalov favoured at 1.69.
Independent estimates of who wins, each from a different method. The blend is what the bot prices against the market.
| Method | A. Gea | D. Shapovalov |
|---|---|---|
| Elo ratings-based, surface-weighted |
51.4% | 48.6% |
| Machine learning 29-feature gradient-boosted model |
49.3% | 50.7% |
| Serve / return point-by-point simulation |
59.3% | 40.7% |
| Blend what the bot actually prices off |
50.1% | 49.9% |
| Market implied de-vigged from the opening odds |
42.5% | 57.5% |
We publish how well these score against the closing market, including where they lose: how accurate are tennis predictions?
| Player | Surface | Elo | Record | Matches |
|---|---|---|---|---|
| A. Gea | clay | 1586 | 72–67 | 139 |
| D. Shapovalov | hard | 1830 | 102–78 | 180 |
Our own ratings, computed from match results since 2018. Players with fewer than 5 rated matches are omitted rather than shown at a provisional starting value. Last updated May 24, 2026.
A. Gea leads 1–0 over 1 meeting. Full record.
| Date | Event | Surface | Winner | Score |
|---|---|---|---|---|
| Aug. 2, 2026 | Los Cabos | hard | A. Gea | 2 - 0 |
| W | D. Shapovalov | 2 - 0 | Aug. 2, 2026 |
| W | C. Wong | 0 - 2 | Aug. 1, 2026 |
| W | F. Cerundolo | 1 - 2 | July 31, 2026 |
| W | M. Zheng | 2 - 1 | July 30, 2026 |
| W | S. Kwon | 1 - 2 | July 29, 2026 |
| L | A. Gea | 2 - 0 | Aug. 2, 2026 |
| W | C. Norrie | 1 - 2 | Aug. 1, 2026 |
| W | D. Svrcina | 2 - 0 | July 31, 2026 |
| W | R. Pacheco Mendez | 2 - 0 | July 30, 2026 |
| W | R. Hijikata | 2 - 1 | July 29, 2026 |
All Aug. 2, 2026 results & picks →
A. Gea (21, ATP 241) vs D. Shapovalov (27, ATP 37) - Los Cabos FINAL, hard.
MODELS vs MARKET
ELO 51.4pct | ML 49.3pct | ensemble 50.1pct (all Gea) vs Betfair-implied ~43pct (2.32).
Nominal Gea edge 4.9pct (ML) to 7.1pct (ELO), ensemble ~5.8pct. Real gap but marginal.
ELO/RANK SANITY
Gea ELO-favoured yet ranked 241 - classic HIGH-ELO / MODEST-RANK flag. Baseline is a
Challenger-level player (Granby/Newport/Ilkley, sub-1700 fields, mixed results). ELO
partly farmed, so discount it. Live form partly validates (this week beat Cerundolo 1856,
Zheng 1796, Kwon 1785, Wong 1735 at ATP level) so not dismissed, but the credible number
is ML about 49pct (edge 4.9pct, below the 5pct bar), not ELO 51pct.
RECENT FORM / PATH
Gea: 4 wins deep at THIS venue this week, maximally match-sharp; real wins over strong
fields. First ATP-250 final (career step-up, big level jump vs pedigree). Set-score
orientation is scrambled in the feed so no reliable scoreline/margin read.
Shapovalov: also 4-0 this week, dominant (Svrcina 6-4 6-2, Pacheco 6-3 6-2) plus 3-set
wins over Norrie 1852 and Hijikata, sharper and higher-caliber, the sharper name in a final.
FATIGUE (Step 2)
Both won 3-set SFs yesterday, so the minus-4.5pp hangover haircut applies to BOTH and cancels.
Gea far heavier cumulative load: 696 min, 18 sets, back_to_back 4 (5th match day) vs
Shapovalov 428 min, 12 sets, back_to_back 3. Gea hits an extreme spot (5 straight days).
Softened one tier by age 21 (recovers fast) but still a real disadvantage on the backed
side. Cumulative-minutes gap alone is context, not an edge.
NARRATIVE
No let-down. Venue: both perfect this week. Surface: both on hard for weeks, no transition.
Nothing amplifying a Gea edge.
ALT MARKETS
Handicap: model cover prob null and set data scrambled, no reliable margin read, stay ml.
Totals: line 22.5, model over 47.9pct (edge -2.9pct), under 52.1pct (edge -1.3pct). Both neg.
AH: A plus1.5 at 2.02 (Gea ML not staked; noted for shadow record).
alt_markets present at analysis time.
GATES
G1 edge-origin: PASS. Edge exists numerically pre-fatigue, BUT once ELO is discounted for
farmed rating the credible (ML) edge is 4.9pct, below the 5pct bar.
G2 consensus: PASS(tier). Models cluster about 50pct, but the informative ML sits at/below
break-even; lone-ELO support only, would cap stake 0.5-1.0u even if bet.
G3 fatal-risk: n/a (ML). Self-consistency: cannot back the heavier-fatigued side (Gea,
5th day, 696 min) whose only clean edge comes from a discounted ELO.
VERDICT: PASS. The model-market gap is real but marginal and driven by a farmed-ELO number;
the credible ML edge (4.9pct) is below the 5pct bar. Qualitative factors (final step-up for
a 241-ranked player, Gea extreme cumulative load vs fresher/sharper Shapovalov, sharp
exchange price on Shapovalov) all argue the market is right rather than validating the model.
Backing Gea manufactures an edge; backing Shapovalov means betting against all three models
and fails his min-odds (need about 60pct, priced about 1.71). No priceable edge anywhere, PASS.